Reference image

Reference image

To create a video like this, we have two options: a free method with results that aren't guaranteed, and a paid, professional method with more accurate results. In both cases we use the same reference image and the same prompt, only changing the site and model used.
01

The Prompt Used

This is the full prompt we'll use, with the project image as a Reference Image. The idea is a single continuous take with no cuts: it starts with a pencil sketching the building inside the plane, then the camera passes through the window in an FPV drone flight toward the site while the project is built in a continuous time-lapse, orbits around it, and ends with the last facade piece installed in slow motion, so the final shot lands exactly on the framing of the reference image.

Full Prompt
A cinematic 12-second single continuous take, no cuts and no dissolves at any moment, Hollywood-level direction, photorealistic, ultra-detailed. The camera moves continuously at a constant, steady pace from the first frame until the final piece is installed; all slow motion and time-remapping happen only at the end. The building in the reference image is the main character: its architecture, massing, facade materials, colors, and surroundings must match the reference exactly once completed. The reference image is the final frame of the film. The outdoor lighting, sky, and time of day throughout the entire exterior part of the film are consistent with the lighting of the reference image. The building reaches its final state only through visible, continuous construction, never through a transition.
CONCEPT: "From the sketch to the site." An architect on an airplane is sketching the project. He opens the window, and far below, the project's site is in its very first stage of construction within a dense city. The camera passes through the window without stopping and dives toward the site in an FPV drone flight, while the building is constructed in a continuous time-lapse. The last stage of construction happens inside the reference framing: the final piece is installed in slow motion before our eyes, then time returns to normal.
TIMELINE (ONE CONTINUOUS SHOT):
SKETCH (0–1.8s): Extreme close-up of a pencil sketching the exact building from the reference image in a sketchbook, fine graphite lines, soft warm cabin light. A gentle zoom out reveals the architect seated by the airplane window, the window shade closed, quiet cabin atmosphere.
THE WINDOW (1.8–2.9s): His hand slides the window shade up, and daylight floods in. The camera pushes steadily toward the window: an oval airplane window with a soft rounded cream frame, the shade raised at the top, warm dim cabin interior, shallow depth of field on the window frame. Through it, from a high altitude of about 1,500 to 2,000 meters: a vast, dense city seen from above in an urban design perspective, a grid of streets, blocks, and buildings stretching to the horizon, and the project's site as a small plot in the middle of the urban fabric, in its earliest stage: excavation, tiny cranes and machines fixing the foundations.
PASSING THROUGH (2.9–3.3s): The camera's forward push never stops: it keeps exactly the same speed as it passes through the window glass with a subtle lens flare, and immediately tilts downward into the dive. The airplane's wing whips past the frame in a fraction of a second. No pause, no hovering beside the airplane.
FPV DIVE (3.3–6.0s): A long, continuous FPV drone dive from high altitude down toward the site, at a constant speed, smoothly banking, with natural motion blur. The city grows larger and more detailed below as the camera descends. As it dives, the construction time-lapse begins and continues at a steady rate: foundations, then the structure rising floor by floor, cranes rotating, workers as streaks of motion, shadows and clouds moving across the city. No day-night flickering.
ORBIT (6.0–8.0s): The camera reaches the project's level and flies a true FPV orbit around the building: it continuously travels sideways around it at a steady speed, with the building rotating in the frame, skimming past the facades, gently banking into the curve. Smooth, gyro-stabilized flight, no shaking, never hovering in place. Meanwhile, the time-lapse continues: facades, glazing, and cladding are installed, matching the reference, and the tower cranes and scaffolding are dismantled and disappear piece by piece. The orbit gradually curves toward the viewpoint of the reference image and arrives at a slightly closer, tighter version of the reference framing by 8.0s, with the building fully complete except for one missing facade piece clearly visible in the frame.
THE FINAL PIECE (8.0–9.8s): Seen from this near-reference viewpoint, only the crane's cable and hook enter from the top of the frame, lowering the final facade piece toward the gap. As it approaches, time remaps smoothly from the time-lapse into slow motion with an ease-out curve. The piece locks into its exact place, completing the facade exactly as in the reference, construction dust suspended in the air. The empty hook then rises slowly out of the top of the frame. The camera keeps a very slow, gentle forward drift, never stopping.
REAL TIME (9.8–12s): Time returns to natural real-time speed. The dust settles and clears, and light everyday street life begins: a few people walking, cars passing. A slow, gentle zoom out from the tighter framing lands exactly on the composition, angle, and framing of the reference image, which is the final frame of the film. At 9.8s, a tiny airplane silhouette appears very far away, several kilometers behind the building, beyond the distant city skyline, low in the sky near the horizon. It is extremely small in the frame, barely larger than a bird, and never passes in front of, above, or close to the building. It slowly descends to land and disappears behind the distant skyline by 11.4s, so the final frame matches the reference image exactly, with no airplane, crane, or dust in it.
SOUND DESIGN (NO MUSIC): Realistic, layered sound effects only:
- Sketch: soft airplane cabin hum, pencil scratching paper, the rustle of a page.
- Window: the slide of the window shade, a soft swell of air.
- Passing through: a subtle airy shimmer and whoosh synced to the flare, flowing directly into the wind.
- FPV dive and orbit: a steady rushing wind roar, with sharp whooshes as the camera skims past facades and cranes.
- Time-lapse: compressed construction sounds at a steady fast pace: drills, hammers, cranes, trucks, metal clanks.
- Final piece in slow motion: deep, muffled, stretched low-frequency sound, the creak of the crane cable, and a heavy, satisfying stretched clunk as the final piece locks into place.
- Real time: calm street ambience: footsteps, a passing car, a light breeze. Exactly when the speed returns to normal at 9.8s, the distant jet engine sound of the landing airplane rises softly, then slowly fades as it disappears behind the skyline, leaving a faint distant rumble on the final frame.
CAMERA & STYLE: One unbroken camera path from the cabin to the site, gimbal feel inside the cabin, smooth gyro-stabilized FPV drone feel outside, constant speed until the final piece, natural motion blur during the dive and orbit, crisp detail in slow moments, time-remapping only at the end with ease-out curves, cinematic color grading (warm, dim, intimate tones in the cabin; exterior lighting matching the reference image), anamorphic film look, 24fps.
02

Choosing a Method

Whichever site you use, the steps are usually similar: upload the image as a Reference Image, add the prompt, set the appropriate settings, choose the model, then click Generate.

Free, Not Guaranteed

Google Flow

Google Flow is completely free, using the Omni 1.1 Flash model.

Paid, Professional

Seedance 2 / 2.5

A professional model available on many sites, so pick the one you prefer. I personally use OpenArt (gives you a discount when you sign up through it).

03

Generating with OpenArt

  • 1Go to OpenArt and open the Video section.
  • 2Choose the model (Seedance 2.5).
  • 3Add the reference image using the Add visual references option.
  • 4Add the full prompt from section one.
  • 5Choose the right settings: audio (On), dimensions (9:16), duration (12s), and resolution (you can pick a low resolution like 480p to save credits, then Upscale as in the next section).
  • 6Click Generate.
Very Important

The image must be added as a "Reference", not as a "Start Frame". The film starts inside the plane, and the reference image is the final shot the project reaches once it is fully built.

Adding the reference image along with the prompt, and setting the model and options, with a low resolution (480p) to save credits

Adding the reference image along with the prompt, and setting the model and options, with a low resolution (480p) to save credits

04

Saving Credits with Upscale

If you feel the video will cost too many credits, you can generate it at a lower resolution, then Upscale the result. This saves a lot of credits. The Upscale uses an engine called BytePlus, on OpenArt itself.

  • 1After generating the video, click on it inside the site to open its options menu.
The menu that opens after clicking the video, with the Upscale option

The menu that opens after clicking the video, with the Upscale option

  • 2Click Upscale, and the video opens with the Upscale settings.
  • 3Choose the settings that suit you (BytePlus vCube engine, resolution, frame rate, mode), no need to copy mine. The cost is relatively low.
  • 4Click Generate.
My Upscale settings (BytePlus vCube, 1080p HD, 30 fps, Pro mode). Choose what suits you, you don't have to copy them

My Upscale settings (BytePlus vCube, 1080p HD, 30 fps, Pro mode). Choose what suits you, you don't have to copy them

05

The Free Alternative: Google Flow

Important Before You Start

This prompt is written for a 12-second video, while Google Flow supports a maximum of 10 seconds. So copy the same prompt into ChatGPT or Claude, ask it to adjust the prompt to fit a 10-second video, then use the adjusted version.

  • 1Open Google Flow.
  • 2Add the image and the adjusted prompt (using the Ingredients option, not Frames, for the same reason mentioned above).
  • 3Choose the right settings and model (Omni 1.1 Flash), with 9:16 dimensions and a 10s duration.
  • 4Click Generate.
Choosing Ingredients, dimensions (9:16), the model (Omni 1.1 Flash), and duration (10s) on Google Flow

Choosing Ingredients, dimensions (9:16), the model (Omni 1.1 Flash), and duration (10s) on Google Flow

Final Result

One continuous 12-second take: a pencil sketch inside the plane, passing through the window and diving toward the site, the project built in a continuous time-lapse, then the last facade piece installed in slow motion, landing on a final frame that matches the reference image, with realistic sound effects only and no music.

Special Discount

If you'd like to subscribe to OpenArt, the link above gives you a special discount when you sign up through it.

If you are still presenting your projects like it's 2019, this is your sign to stop. Click the link below to learn more.

I want to present better